| Bucket | Condition |
|---|---|
low | 1% ≤ |delta_pct| < 5% |
moderate | 5% ≤ |delta_pct| < 15% |
high | 15% ≤ |delta_pct| < 30% |
hyper | |delta_pct| ≥ 30% |
Why buckets matter
A raw event stream is a firehose. Magnitude buckets turn it into a prioritized queue.Filter the noise
Ignore every
low blip and only review high and hyper events when you are time-constrained.Scale by company size
The percentage basis means a single hire at a 10-person company buckets far higher than one hire at a 5,000-person company, exactly as it should.
Rank your queue
Sort a day’s signals by magnitude to work the strongest events first.
Tune your alerts
Set alert thresholds per signal: maybe
hyper only for headcount, but moderate and up for funding.Categorical and scored signals
Some signals are not percentage-based and are handled differently.Inherently high-signal events
Inherently high-signal events
Events like
employee_size_band_upgrade, first_job_in_function, and name_change_drastic are categorically meaningful and are treated as high-signal regardless of a raw percentage. Crossing a size band, for example, requires sustained net hiring, so it rarely fires on noise.Continuous scores
Continuous scores
momentum_score and risk_score are not bucketed events. They are continuous, decaying scores recomputed nightly, designed to be sorted and ranked rather than bucketed. See the composite signals.Browse the full signal index
Every signal, its category, and what it means, in one table.